This guide is for HR and recruiting teams using Tactiq with Claude. It covers what each one is for and the prompts worth saving.

Claude can find your meetings on any plan, but opening them needs Team or above.

Per interview, you want a consistent write-up fast, while the call is still fresh. A saved prompt inside Tactiq serves that better than Claude does, because it runs in one click with no setup.

Across a hiring round, you want to compare candidates who were interviewed separately, see where the panel disagreed, and catch where you asked one person a harder question than another. Spaces will do a lot of this. Claude is for when you want it worded the same way every time, or landing somewhere else.

Use the first for speed. Use the second before you make a decision.

Claude can do five things with your interviews, in the order you will actually use them.

Find the interview. By candidate name, by role, by date range, or just the last few. You do not have to remember which day you spoke to someone.

Read the summary. The richest single view of a call. For most post-interview questions this is enough on its own.

Search inside a call for the exact moment. What did she actually say about notice period. What was the answer on the systems migration question. You get the line and a link back to the point in the recording.

Pull the whole transcript when you genuinely need every word, which in hiring is rarer than you would think. Mostly this is for disputes and appeals.

Pull other artifacts already generated on that meeting, such as action items or an email draft.

Everything comes back with a link to the meeting in Tactiq. When a quote is going to appear in a scorecard or a rejection rationale, click through and check it before it leaves your hands.

Spaces come up throughout, so briefly: a Space is a shared space of meetings with its own page and its own permissions. Put a role's interviews in one and the hiring team can all see them. From there you can select several interviews and ask a question across the set, and the answer carries citations that jump to the moment in a named candidate's call. Worth knowing that adding a meeting to a Space gives everyone in that Space access to it.

Five habits. The first three apply to any meeting. The last two are specific to hiring and they are the ones that matter most.

Search first, read second. Let Claude find the interview by candidate, role or date.

Ask for the moment, not the meeting. Most questions have a one-line answer buried somewhere in the call. Ask for that line and you get it back with a link to the exact point in the recording, so you can check it before repeating it.

This matters most when the wording does. Did she say she would relocate, or that she would consider it? A summary is likely to record both as a yes. You find out which one she meant when you make the offer.

Name people, roles and dates. "My interviews for the senior analyst role in August" resolves cleanly. "That candidate from a while back" does not.

Ask for the evidence, not the verdict. This is the big one. If you ask whether someone is a good fit, you will get a confident answer built partly on how fluent they sounded. Ask instead which requirements have evidence behind them, what that evidence was, and what was never tested. You stay the one deciding, and you end up with a rationale you can show someone.

Ask what was missed. A summary tells you what was said, not what nobody got round to asking. So ask for it: "list anything in the job description that never came up." Those gaps are your questions for the next round.

Most of these run in Claude. Where Tactiq does the job as well or better, it says so.

Before the next round. Pull the earlier interviews with this candidate and get what was already covered, what they asked us, and what is still unproven. Five minutes of this and you stop asking them why they are leaving their current job for the third time.

Straight after a call. The exception: this one is Tactiq's. Run your scorecard and candidate details from a saved kit while the call is fresh. One click, no setup, and the same structure for every candidate. Go a step further and set it up as a workflow, so it runs on its own after every interview and pushes the output into wherever your team actually works.

Comparing a shortlist. Either works. A Space is the quicker route: put the round's interviews in one, select them, ask one question across the set, and you get an answer you can push back on with citations that jump to the moment in a named candidate's interview.

Go to Claude when the comparison needs to meet something outside Tactiq. It can connect to your ATS, so it can read the interviews and write the result back where the rest of the hiring record lives.

Checking your own interviewing. Ask what questions you ask everyone versus only some people, and whether you probe harder with certain candidates. Uncomfortable and worth doing. It is also the strongest answer you have if anyone ever questions how a decision was reached.

Panel disagreement. When two interviewers land in different places, pull both transcripts and ask where the evidence actually diverges. Often, they asked different questions rather than reading the same answer differently.

Hiring manager updates. Pull the round, produce the state of play: who is moving, who was declined and why in one line, what is blocking a decision.

Closing the loop on a rejection. Before sending feedback, pull the specific moments behind the decision so what you tell the candidate matches what actually happened in the room.

Swap the bracketed parts. These are written to be copied.

Prep for the next round

Pull my previous interviews with [candidate] for [role]. Tell me what was already covered, what they asked us, what we committed to, and which requirements are still untested. Keep it under 200 words.

Scorecard, one candidate

Find my [date] interview with [candidate] for [role]. Score them against each requirement in the job description below. For each one: the requirement, a rating of Strong evidence / Some evidence / Not covered / Contradicting evidence, and the exact quote behind it. Rate only on what was said, never on tone or confidence. Where something never came up, say not covered rather than guessing. Then list the untested requirements as questions for the next round.

Job description: [paste]

Compare the shortlist

Find every interview for [role] between [date] and [date]. Score each candidate against the same requirements, worded identically. Give me a table with candidates down the side and requirements across the top.

Where a requirement never came up in someone's interview, leave that cell empty rather than rating it. Do not infer a rating from something they said about a different topic.

Quote the line behind every rating. Under the table, list the requirements that were not put to every candidate.

Panel disagreement

Pull the interviews with [candidate] by [interviewer A] and [interviewer B]. Where does the evidence actually diverge, and where did they simply ask about different things?

Calibrate yourself

Find every interview I have run for [role] in the last [n] months. Which questions do I ask everyone, and which only sometimes? Do I probe harder with some candidates than others? Anywhere I raised a concern in a debrief that the transcript does not support? Be blunt.

Hiring manager update

Pull every interview for [role] this month. Write the update: where the pipeline stands, who is moving, who was declined with a one-line reason, and the two things blocking a decision.

Before sending feedback

Find my interview with [candidate] and give me the specific moments behind our decision, with quotes, so the feedback I send matches what actually happened.

Asking whether they are any good. A transcript is only words, so fluency is most of what there is to judge. Ask that question, and you get an answer shaped partly by how articulate someone sounded, which lands hardest on people interviewing in a second language or anyone who was nervous. Ask for evidence against requirements.

Asking across a Space without selecting the meetings. Hit Ask AI with nothing ticked, and it picks the top three for you. It says so, but it is easy to miss, and a confident comparison of three candidates out of five looks exactly like a comparison of five. Select every interview you mean to include, every time.

Comparing candidates who were not asked the same things. A table gives everyone a rating in each row, even where the question came up in only two of the five interviews. Tell Claude to leave the cell empty when the question was never asked, instead of rating it.

Once a prompt works well, save it rather than rewriting it next time.

Save the per-interview prompts as a kit. Your scorecard, your candidate-details extraction, your debrief note. Saved in Tactiq, they run in one click on any interview, with no setup and no connector. This is where most of the volume is, and it is the part your colleagues will actually adopt.

Share the kit, not the prompt. Share it with your team, and everyone runs the same one. A prompt pasted into Slack gets used once by one person. A shared kit gets used every week by everyone. It is also how you stop five recruiters writing five slightly different scorecards and then trying to compare candidates across them.

Save the round-level ones as a Claude skill. Comparisons, calibration, panel disagreements, and hiring manager updates. Same principle as the kit: written once, reused every round, identical wording every time. That last part is what makes candidates comparable, because a scorecard you reworded between candidate two and candidate five is not a comparison.

The rule of thumb: if you do it after every interview, it belongs in a kit. If you do it once per hiring round, it belongs in Claude.